AN ADAPTIVE, ON LINE, STATISTICAL METHOD AND APPARATUS FOR MOTOR BEARING FAULT DETECTION BY PASSIVE MOTOR CURRENT MONITORING

摘要

A motor current signal is monitored during a learningstage and divided into a plurality of statisticallyhomogeneous segments representative of good operatingmodes. A representative parameter and a respectiveboundary of each segment is estimated. The current signalis monitored during a test stage to obtain test data, and thetest data is compared with the representative parameterand the respective boundary of each respective segment todetect the presence of a fault in a motor. Frequencies atwhich bearing faults are likely to occur in a motor can beestimated, and a weighting function can highlight suchfrequencies during estimation of the parameter. Thecurrent signal can be divided into the segments by dividingthe current signal into portions each having a specifiedlength of time; calculating a spectrum strip for eachportion; and statistically comparing current spectra ofadjacent ones of the strips to determine edge positions forthe segments. Estimating the parameter and the boundary ofeach segment can include calculating a segment mean (therepresentative parameter) and variance for each frequencycomponent in each respective segment; calculating amodified Mahalanobis distance for each strip of eachrespective segment; and calculating the modifiedMahalanobis mean and the variance for each respectivesegment. Each modified Mahalanobis mean can form arespective radius about a respective segment mean todefine a respective boundary.